The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Data » Dataset use in Stata » Merging Datasets and Matching Observations BDHS 2007 (Domestic Violence and Decision Making Variables)
Merging Datasets and Matching Observations BDHS 2007 [message #16657] Wed, 13 February 2019 05:41 Go to next message
leyfh1 is currently offline  leyfh1
Messages: 3
Registered: November 2018
I am using the Bangladesh DHS from 2007.
I would like to run regressions including the domestic violence variables and the variables regarding different decisions a woman can make (based on her health, family visits, large household purchases etc.). However, the variables regarding decision making are in the couple's dataset and the variables regarding domestic violence are in the woman's dataset. When I merge the 2, the women who answered questions on domestic violence did not answer questions regarding decision making and vice versa so I am unable to test my hypothesis. I have read many studies where this dataset and the variables in question have been used together.

Could you please advise me on how I go about solving this problem?
What could I be doing wrong?
Is there something I need to be aware of when merging these datasets?
How do I match the women in the couple's questionnaire to the women in the women's questionnaire?

Many thanks!
Re: Merging Datasets and Matching Observations BDHS 2007 [message #16689 is a reply to message #16657] Mon, 18 February 2019 08:49 Go to previous message
Bridgette-DHS is currently offline  Bridgette-DHS
Messages: 1977
Registered: February 2013
Senior Member

Following is a response from Senior DHS Specialist, Kerry MacQuarrie:

For the analysis you want to do, you should be working entirely within the woman's individual recode (IR) file. You should not need to use or merge with the couples (CR) file.

In the IR file, the decision-making variables that you seek are:
v743a--final say on own health care
v743b--final say on large household purchases
v743c--final say on household purchases for daily needs
v743d--final say on visits to family or relatives
(There is also v739--who decides how to spend money and s826--who usually makes decisions about child's health care.)

These decision-making variables are answered by all 10,996 women in the sample because this survey is an ever-married women survey and these questions are only relevant for partnered women. (In DHS surveys of all women, the decision-making variables are only answered by currently or formerly married women.)

The domestic violence module was answered by only 4,467 women, all of whom have valid response on the decision-making variables. So this is what your analytic sample will be. The smaller sample for the domestic violence module is because for ethical reasons we only interview one woman per household with the violence module, whereas all eligible women in the household complete the main questionnaire. Please remember to use d005 (the domestic violence weight) as the basis of your weight variable rather than v005.

In the case of the Bangladesh 2007 survey, the survey had two sub-samples, with the violence module implemented in one sub-sample and the men's questionnaire implemented in the other sub-sample. This is why, when you looked at the CR file, you found no women who had completed the domestic violence questions.

I hope this helps!
Previous Topic: Multilevel modeling in DHS-Sri Lanka
Next Topic: Error with Stata do-file: DAT file called
Goto Forum:

Current Time: Wed Oct 28 07:16:30 Eastern Daylight Time 2020